feat: Update llama.cpp

abetlen · May 8, 2024 · 2a39b99 · 2a39b99
1 parent 9ce5cb3
commit 2a39b99
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 3 deletions.
diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py
@@ -296,7 +296,9 @@ def byref(obj: CtypesCData, offset: Optional[int] = None) -> CtypesRef[CtypesCDa
 #     LLAMA_VOCAB_PRE_TYPE_GPT2           = 7,
 #     LLAMA_VOCAB_PRE_TYPE_REFACT         = 8,
 #     LLAMA_VOCAB_PRE_TYPE_COMMAND_R      = 9,
-#     LLAMA_VOCAB_PRE_TYPE_OLMO           = 10,
+#     LLAMA_VOCAB_PRE_TYPE_QWEN2          = 10,
+#     LLAMA_VOCAB_PRE_TYPE_OLMO           = 11,
+#     LLAMA_VOCAB_PRE_TYPE_DBRX           = 12,
 # };
 LLAMA_VOCAB_PRE_TYPE_DEFAULT = 0
 LLAMA_VOCAB_PRE_TYPE_LLAMA3 = 1
@@ -308,7 +310,9 @@ def byref(obj: CtypesCData, offset: Optional[int] = None) -> CtypesRef[CtypesCDa
 LLAMA_VOCAB_PRE_TYPE_GPT2 = 7
 LLAMA_VOCAB_PRE_TYPE_REFACT = 8
 LLAMA_VOCAB_PRE_TYPE_COMMAND_R = 9
-LLAMA_VOCAB_PRE_TYPE_OLMO = 10
+LLAMA_VOCAB_PRE_TYPE_QWEN2 = 10
+LLAMA_VOCAB_PRE_TYPE_OLMO = 11
+LLAMA_VOCAB_PRE_TYPE_DBRX = 12
 
 
 # // note: these values should be synchronized with ggml_rope
@@ -377,6 +381,7 @@ def byref(obj: CtypesCData, offset: Optional[int] = None) -> CtypesRef[CtypesCDa
 #     LLAMA_FTYPE_MOSTLY_IQ2_M         = 29, // except 1d tensors
 #     LLAMA_FTYPE_MOSTLY_IQ4_XS        = 30, // except 1d tensors
 #     LLAMA_FTYPE_MOSTLY_IQ1_M         = 31, // except 1d tensors
+#     LLAMA_FTYPE_MOSTLY_BF16          = 32, // except 1d tensors
 
 #     LLAMA_FTYPE_GUESSED = 1024, // not specified in the model file
 # };
@@ -409,6 +414,8 @@ def byref(obj: CtypesCData, offset: Optional[int] = None) -> CtypesRef[CtypesCDa
 LLAMA_FTYPE_MOSTLY_IQ2_S = 28
 LLAMA_FTYPE_MOSTLY_IQ2_M = 29
 LLAMA_FTYPE_MOSTLY_IQ4_XS = 30
+LLAMA_FTYPE_MOSTLY_IQ1_M = 31
+LLAMA_FTYPE_MOSTLY_BF16 = 32
 LLAMA_FTYPE_GUESSED = 1024
 
 # enum llama_rope_scaling_type {

diff --git a/vendor/llama.cpp b/vendor/llama.cpp
+4 −0		convert-hf-to-gguf-update.py
+6 −0		convert-hf-to-gguf.py
+32 −19		convert.py
+1 −1		examples/finetune/finetune.cpp
+2 −1		examples/quantize/quantize.cpp
+1 −1		examples/server/README.md
+2 −1		examples/server/server.cpp
+13 −1		examples/server/tests/features/server.feature
+32 −4		examples/server/tests/features/steps/steps.py
+12 −12		examples/server/utils.hpp
+77 −0		ggml-impl.h
+1 −1		ggml-metal.m
+18 −0		ggml-quants.c
+1,089 −88		ggml.c
+15 −7		ggml.h
+2 −0		gguf-py/gguf/constants.py
+32 −2		llama.cpp
+4 −1		llama.h
+ −		models/ggml-vocab-qwen2.gguf
+106 −0		models/ggml-vocab-qwen2.gguf.inp
+43 −0		models/ggml-vocab-qwen2.gguf.out
+9 −6		scripts/compare-llama-bench.py
+1 −0		tests/CMakeLists.txt
+4 −2		tests/test-backend-ops.cpp